بیماری آلزایمر با از دست دادن ویژگی های متمایز معناشناختی همراه است

A central topic of discussion in the exploration of semantic disturbance in Alzheimer's disease (AD) concerns the relative contribution of semantic content (e.g., semantic features) and semantic process. Studies have suggested that semantic dysfunction in AD is the result of deficits to either semantic process, semantic content or both. Studies that have supported the loss of semantic content have been criticised for their use of verbal stimuli and cognitively challenging experimental tasks. The current study used a novel version of the yes–no recognition memory task to compare the processing of distinctive and non-distinctive features in participants with AD whilst controlling the cognitive demands of the task. The task involved five conditions which denoted the relationship between the items in the test and study phase. A ‘non-distinctive’ and a ‘distinctive’ condition were included where non-distinctive and distinctive semantic features were manipulated between study and test, respectively. Task accuracy of participants with AD decreased relative to control participants when distinctive features were manipulated between the study and test phase of the experiment. There was no significant difference between groups when non-distinctive features were manipulated. These findings provide evidence to support the loss of semantic content in AD.

1.1. Semantic disturbance in AD Language difficulties are very common for people with Alzheimer's disease (AD) (Cummings, Benson, Hill, & Read, 1985). The key clinical language characteristics of people with AD are word finding difficulties, semantic paraphasias and comprehension deficits with the relative sparing of repetition, morphology, syntax, phonology and motor speech skills (Blair et al., 2007 and Huff et al., 1986). It is widely regarded that the language deficits observed in AD stem from semantic disturbance (Chenery, 1996, Christensen et al., 1998, Garrard and Carroll, 2006, Huff et al., 1986 and Martins and Lloyd-Jones, 2006). As a consensus has not been reached in the literature as to the nature of semantic disturbance in AD, this study aims to explore competing accounts. Alzheimer's disease is associated with multiple cognitive deficits and it is widely accepted that the semantic disturbance of AD involves deficits in semantic control (Reilly, Peelle, Antonucci, & Grossman, 2011). Semantic control is a collective term for the cognitive processes and skills believed to be necessary for completing semantic tasks and includes attention, memory, noise suppression, and enactment (Jefferies, Patterson, & Ralph, 2008). Of controversy in the literature exploring language in AD is the co-occurrence of disturbance to underlying semantic representations (Hornberger, Bell, Graham, & Rogers, 2009). The debate echoes Warrington and Shallice's (1979) access versus storage dichotomy where a division is drawn between semantic deficits caused by either disturbances to the access of semantic information or disturbance to the semantic information itself. However, the discussion of semantic disturbance could be more accurately described as one of access and storage (or process and content) versus access only (Reilly et al., 2011). Authors who have supported the loss of semantic content have claimed that the neuropathology of AD compromises structures responsible for storing semantic information, and loss of this information contributes to semantic disturbance (Chertkow and Bub, 1990 and Martin, 1992). Authors supporting the loss of semantic process only have suggested that semantic information remains largely intact but access to semantic information is impeded by the cognitive decline associated with the disease (Daum et al., 1996, Hartman, 1991, Nebes et al., 1989 and Nebes et al., 1984). Each side of the debate is based upon a different theory of semantic representation and, therefore, the resolution of this argument has implications for our understanding of semantic disturbance in AD as well as our understanding of the structure of the semantic system. In Warrington and Shallice's (1979) discussion of access versus storage, storage/content disturbances were associated with consistent errors, frequency effects, selective loss of subordinate information, lack of priming and cueing effects. These criteria have been criticised for their lack of empirical or consistent theoretical basis (Rapp & Caramazza, 1993). Many subsequent studies that have supported loss of content in AD have focused upon semantic feature loss (Almor et al., 2009, Garrard et al., 2005 and Laisney et al., 2011). Semantic features can be defined as a description of the properties of a concept and can be classed as distinctive (defining features and therefore common to only one or two concepts) or non-distinctive (non-defining features and therefore common to many concepts) (McRae, Cree, Seidenberg, & McNorgan, 2005). A number of computational models have been developed to explain semantic disturbance in AD that rely on semantic feature loss. Prominent models include Gonnerman, Andersen, Devlin, Kempler, and Seidenberg (1997), Plaut (1996) and Develin, Gonnerman, Andersen and Seidenberg (1998). Within these models, concepts are generally viewed as an interconnected network of semantic features. As non-distinctive features are common to many items within a category, they are frequently co-active, highly intercorrelated and therefore share numerous, strong links. Distinctive features, on the other hand, share few interconnecting links. With cortical damage characteristic of AD, interconnections between semantic features are expected to be lost. Being highly intercorrelated, non-distinctive features are not affected by the loss of interconnections early in the disease as many other connections are available to compensate. Distinctive features, however, are vulnerable early in the disease process. Although these connectionist models were originally used to describe category specific deficits in AD, the vulnerability of distinctive semantic features has also been used to explain other patterns of semantic error in AD (Done and Gale, 1997, Garrard et al., 2005, Laisney et al., 2011 and Rogers and Friedman, 2008). Early in the disease, people with AD are frequently reported to make category co-ordinate (e.g., calling a ‘horse’ a ‘cow’) and superordinate (e.g., calling a ‘horse’ an ‘animal’) semantic paraphasias (Bowles et al., 1987 and Hodges and Patterson, 1995). Such errors may be explained by the loss of distinctive semantic features, which make it difficult to distinguish between concepts within a category, yet category level knowledge is retained (Done and Gale, 1997, Garrard et al., 2005, Laisney et al., 2011 and Rogers and Friedman, 2008). In the latter stages of the disease, people with AD are reported to frequently make unrelated errors or give ‘I don't know’ responses on naming tasks (Bowles et al., 1987 and Hodges and Patterson, 1995). These errors may be explained by the loss of numerous interconnections later in the disease process, which compromises non-distinctive features and leads to the loss of category level knowledge (Done and Gale, 1997, Garrard et al., 2005, Laisney et al., 2011 and Rogers and Friedman, 2008). The loss of semantic features in AD, particularly the loss of distinctive features, has been supported by a number of behavioural studies over the past decade (Done and Gale, 1997, Garrard et al., 2005, Laisney et al., 2011 and Rogers and Friedman, 2008). Garrard et al. (2005) requested participants with AD to list as many features as possible for a given object. Compared with controls, participants with AD listed fewer features, with particular difficulty listing distinctive features. The authors also found that the number of features given for an item was correlated with the ability to name that item. Laisney et al. (2011) created a semantic priming experiment where prime-target pairs consisted of concept-attribute pairs of either distinctive (e.g. zebra-stripes) or non-distinctive features (e.g. duck-feathers). Semantic priming was impaired in AD for distinctive attribute-concept pairs but not for non-distinctive attribute-concept pairs, whereas controls showed priming for both conditions. The findings of these studies have led the authors to conclude that semantic disturbance in AD involves the loss of semantic feature information, with distinctive semantic features being more vulnerable to loss than non-distinctive features (Garrard et al., 2005 and Laisney et al., 2011). This differential loss of distinctive semantic features in AD is central to claims of semantic content loss as it is used to explain the progressive nature of semantic disturbance in the disease (Done and Gale, 1997, Garrard et al., 2005, Laisney et al., 2011 and Rogers and Friedman, 2008). A number of authors have suggested that semantic disturbance in AD is due to the disturbance of semantic processes only and not a deficit in semantic representations (Daum et al., 1996, Fung et al., 2000, Nebes et al., 1984, Ober and Shenaut, 1999 and Perri et al., 2003). These authors have claimed that because people with AD have impaired cognition, especially attention and executive functioning, the ability to access otherwise spared semantic representations is also impaired. Furthermore, if the cognitive demands of a task are reduced (in a situation where automatic rather than effortful retrieval can occur), then the semantic disturbance observed in AD would be reduced (Daum et al., 1996, Nebes et al., 1984 and Perri et al., 2003). Evidence for this account comes from a number of behavioural studies exploring semantic functioning in AD while controlling the cognitive demands of the task. The first behavioural evidence for the semantic process only view was published by Nebes et al. (1984), who tested semantic memory in participants with AD and healthy controls using tasks with low cognitive demands (semantic priming, approximation to English and approximation to text tasks). The authors reported that participants with AD performed as well as controls on measures of semantic memory. Similarly, Daum et al. (1996) reported that participants with AD showed improved performance on semantic tasks with low cognitive demands such as object decision and preference judgment tasks compared with effortful tasks such as confrontation naming and definitions. These findings led the authors to reason that the semantic deficits of AD were the product of effortful semantic tasks with high cognitive demands preventing access to intact semantic representations (Daum et al., 1996 and Nebes et al., 1984). Since these early studies, other authors have also reported that the semantic deficits of AD can be reduced through using non-effortful cognitive tasks (Fung et al., 2000 and Perri et al., 2003). Though studies supporting deficits to semantic process only are in the minority, they highlight the importance of minimising and/or accounting for the cognitive demands associated with task performance in AD. 1.2. Theories of semantic representation in relation to AD The outcome of the debate over loss of content in semantic dysfunction in AD has implications for models of semantic representation. Loss of semantic content, unlike disturbance to semantic process only, is consistent with models of semantic representation that rely on grounded representation. Proponents of grounded models assert that semantic memory is based on sensory representations (Barsalou, 2008). These models are distinct from amodal and hub and spoke models in which semantic memory is based on a central store of amodal symbols (Caramazza et al., 1990, Mahon and Caramazza, 2008 and Reilly and Peelle, 2008). Within the context of grounded models, semantic concepts are represented as a distributed network of perceptual symbols/semantic features spread throughout the cerebral cortex, in particular, the somatosensory areas of the brain (Barsalou et al., 2003, Coltheart et al., 1998, Dove, 2009 and Mayberry et al., 2011). Grounded models would predict that because AD involves progressive neural loss in many regions over the cerebral cortex (Perl, 2010), semantic feature loss, or disruptions to interconnections between semantic features, is likely to occur and this progressive loss of semantic features will result in semantic disturbance (Done and Gale, 1997 and Done and Hajilou, 2005). Furthermore, grounded models predict the differential loss of distinctive over non-distinctive semantic features in AD due to greater intercorrelation and stronger synaptic links for non-distinctive features than distinctive semantic features (Done and Gale, 1997, Pulvermuller, 2001 and Ursino et al., 2011). These predictions are consistent with disturbance to semantic process and content, but not process only, and therefore the debate between these competing accounts has implications for semantic theory. Thus, if an experiment with reduced cognitive demands reveals a processing deficit for semantic features in AD relative to control participants, this would provide evidence for semantic content loss and grounded modelling. 1.3. Semantic tasks and cognitive demands One of the key limitations of previous studies that have found evidence for loss of semantic content in AD is the use of verbal stimuli. More specifically, the difficulties that people with AD have in processing verbal stimuli could either mask or exaggerate changes to underlying semantic representations (Garrard et al., 2005). Difficulties processing verbal stimuli do not necessarily reflect changes to the semantic system as words and underlying conceptual representations share only arbitrary links (Mayberry et al., 2011 and Rogers et al., 2004). It is widely held that the processing of lexical stimuli (e.g., reading and naming) can be impaired in the absence of semantic disturbance (Glosser and Friedman, 1991 and Kay et al., 1996). Difficulty processing lexical stimuli could prevent a semantic deficit being detected or be incorrectly interpreted as a semantic deficit. Studies of semantic processing in populations such as AD, could therefore benefit from avoiding lexical stimuli in order to minimise the potential confounds of lexical processing. Picture stimuli may provide a suitable alternative to lexical stimuli and may help lower cognitive confounds in experimental tasks for an AD population. According to the picture superiority effect, people with AD perform better on cognitive tasks using pictures rather than words (Stopford, Thompson, Neary, Richardson, & Snowden, 2012). Moreover, people with AD perform better on semantic tasks using coloured pictures and photographs than the black and white line drawings that are often used in semantic studies (Zannino et al., 2007 and Zannino et al., 2010). Taken together, the use of photographs as stimuli, without the requirement of naming, could potentially reduce the cognitive demands on participants with AD compared with lexical stimuli and provide a more valid assessment of semantic memory. Another advantage to utilising picture stimuli is that it allows for semantic feature processing to be explored through the direct manipulation of semantic features. For example, according to published semantic feature norms (McRae et al., 2005), a semantic feature of ‘deer’ is that they 〈have antlers〉; this feature could be altered or removed when using picture stimuli and the processing of the stimulus compared between participants with AD and controls. Therefore, by selectively altering distinctive and non-distinctive semantic features of picture stimuli, the differential processing of these feature types in AD could be tested. Imaging studies have shown that people with AD have lesions throughout motor and sensory areas of the cerebral cortex with the visual cortex being affected in the early stages of the disease (Frisoni, Prestia, Rasser, Bonetti, & Thompson, 2009). Therefore, if semantic feature loss has occurred, it would be likely to involve distinctive visual semantic features and affect the processing of picture stimuli. The cognitive confounds of previous studies that have found evidence of semantic content loss could be further reduced by using picture stimuli in a simple cognitive task such as the yes–no recognition memory task (Daum et al., 1996 and Nebes et al., 1984). The yes–no recognition memory task involves a participant being shown a number of items during a learning phase, and then being re-shown these ‘old’ items with the addition of previously unseen ‘new’ items in a test phase (Macmillan & Creelman, 2005). The participant must learn the old items in the study phase and correctly determine which items are old and those that are new in the test phase. Performance is generally reported in terms of the ‘hit rate’ (proportion of correct response to old items) and ‘false alarm rate’ (proportion of incorrect responses to new items) (Macmillan and Creelman, 2005 and Snodgrass and Corwin, 1988). The yes–no recognition memory task has been previously used to explore semantic processing in AD by using semantically related and unrelated new items. Budson, Daffner, Desikan, and Schacter (2000) found that false alarm rates increased for control participants but not participants with AD when stimuli in the test phase were semantically related to items in the study phase compared to items that were unrelated. The novel use of this paradigm involving the manipulation of semantic features of stimuli could provide new insight into semantic feature processing in AD and the presence of disturbance to semantic content. The present study aimed to investigate disturbance to semantic content in AD using a simple cognitive task to minimise the potential confounds of semantic process. Many contemporary models have used the semantic feature as a basic unit of semantic content and a number of behavioural studies have cited the loss of semantic features as a contributing factor to semantic disturbance in AD. To this end, distinctive and non-distinctive semantic features of picture stimuli were manipulated between study and test phases of a yes–no recognition experiment. Five conditions were included: the identical condition (old items and therefore identical between study and test), the non-distinctive condition (new items altered by non-distinctive features between study and test), the distinctive condition (new items altered by distinctive semantic features between study and test), the semantic condition (new items that are category co-ordinates of items in the study phase) and the unrelated condition (new items that are not related to items in the test phase). Based on the notion that semantic feature loss may contribute to semantic disturbance in AD, and that non-distinctive features are more resistant to loss in the disease, a differential performance across groups was predicted. Specifically, it was predicted that if a non-distinctive feature was altered between study and test phases, participants with AD and control participants would have similar false alarm rates. When distinctive features were altered between study and test phases, then participants with AD were predicted to have a significantly higher false alarm rate than control participants.